wewbull

joined 3 years ago
[–] wewbull@feddit.uk 2 points 3 days ago

It's a term most associated with Steve Jobs back in the day.

[–] wewbull@feddit.uk 5 points 3 days ago* (last edited 3 days ago) (2 children)

I wasn't questioning STEM.

Op mentioned consumer OS market share stats, and the only ones I know of are from Steam.

[–] wewbull@feddit.uk 7 points 3 days ago (1 children)

I had an argument with my boss where he was trying to stop people checking out source code to company laptops because they could be lost. My retort was that the company uploaded all it's company documents, spreadsheets, and other data to the cloud including employee personal data, and that was far for likely to experience a data breach. Those breaches would result in legal issues for the company, rather than a slight embarrassment.

He was apoplectic. How dare I question company policy.

[–] wewbull@feddit.uk 9 points 3 days ago (6 children)

Do you mean steam stats? I don't know of many others, and those are going to be enthusiasts of one form or another.

[–] wewbull@feddit.uk 23 points 5 days ago

It has some reasonable use-cases when dealing with text

[–] wewbull@feddit.uk 3 points 5 days ago* (last edited 5 days ago)

I think that these days centres should absolutely not be built. They are on a scale we can't justify economically or ecologically and the technology they support are national security threats.

These are not the AWS-like data centres that "host the internet". These are compute only. They're for AI and big-data. Centralisation of the world computation into a few locations under the control of a few billionaires.

So point 1. Destroying these does not harm the internet. It preserves it. The 90s internet was one where anyone could put up a web-server and serve whatever they wanted. Sure, it might get given a hug-of-death when a site became popular, but it was only that one site that was affected. The internet was distributed and not centralised.

Then came the cloud.

Point 2. This is centralising information. I suspect we are already past the point where people are asking chatbots "who should I vote for?". That alone places millions of votes under the control of whoever trains the LLM model they ask.

The owners become the owners of all information. It's so handy. Why read a book? In fact let's destroy books to create the machine. Now you can ask it anything. How do I make a cake? How do I fix my car? Is there a war in ~~Palestien~~ Ba-sing-se?

Point 3. The cost is too high. The American stock market is teetering on the edge of a huge crash. Bigger than 2008, it threatens to take out several of the world's most valuable companies. If the damage was limited to them, I wouldn't worry. Sadly, the banks and investment funds we all use are invested in them. This is Enron x4 plus the banking failure of 2008 in a dot.com bubble wrapper. All whilst the economy in all other sectors is shrinking.

Wall street knows it's a bubble, but they all want to ride the upside of the bubble, thinking it won't be them left holding when the crash comes. They could calmly unwind their positions and avoid the crash, but ego and hubris don't allow that.

All in pursuit of an impossible goal

Then there's the ecological cost. Lake Mead, on the Colorado river, is at it's lowest level since it was created by the Hoover Dam being constructed. Drought and water use in the south-western states have steadily drained it for years ...and they want to add a water guzzlers on the scale of an additional cities. Now I know this one is in Texas, and not in the Colorado river basin, but this isn't the only one and I'm sure Texas rivers have their own issues.

Then you have the power supply. Whilst the rest of us are trying to move away from fossil fuels, these guys run towards them. Natural gas turbines (which can't be built to schedule) belching out CO2 whilst heatwaves, drought, and wild fires are happening across the globe. I guess it gives the fossil fuel companies a reason to keep going though.

So yes, I'm against building these centres. The benefits are low, the societal dangers are huge and...

[–] wewbull@feddit.uk 1 points 6 days ago (1 children)

You've seen where they want to build this, right? Heat is not an issue.

[–] wewbull@feddit.uk 1 points 6 days ago

So don't build it. It needs too much power.

[–] wewbull@feddit.uk 1 points 6 days ago

I hear jet fuel melts steal beams.

[–] wewbull@feddit.uk 1 points 6 days ago (2 children)

You sound like you acknowledge it's going to cause huge amounts of environmental damage, but because there's no other way of powering it... I guess that's what we're going to have to do.

Is that a fair take on your position?

[–] wewbull@feddit.uk 2 points 6 days ago

Use AI to add the grim reaper into the background?

 

Okay, we have a new contender for Most AI Thing to Ever Happen

  1. July 25th: someone messes around with an LLM and posts a proof of the Collatz conjecture that does, in fact, verify in the theorem prover. (The AI use is not disclosed on the github page) https:// github.com/xrchz/CollatzLean

  2. July 26th: several serious bugs are posted in the theorem provers, that in principle could allow a false statement to be "proven" true. They're serious, yes, but no need for panic, because you're not going to blunder into accidentally exploiting the bugs while writing a proof, probably. https:// github.com/leanprover/lean-kernel-arena/pull/81

  3. July 28th: someone who was right to be very skeptical of the Collatz proof, and had the expertise to study it with a fine-toothed comb, discovered it was exploiting a bug https:// github.com/leanprover/lean4/issues/14576

  4. The "proof" turns out to be exploiting multiple similar but distinct bugs to pass different solver variants!

  5. the human who posted the proof acknowledges the AI use and claims they did not knowingly point it towards the bugs it exploited. https:// leanprover.zulipchat.com/#narrow/channel/270676-lean4/topic/Counterexample.20to.20the.20Lean.20Conjecture.20.28Soundness.20Bug.29/near/613135216

Note that the proof was posted shortly before the related bug reports were posted. It is an open question if the AI found people discussing the bugs shortly before they were formally posted and "decided" to exploit them, if the AI "knew about it" as a learned strategy from the training stage (putting every single "proof" it's ever made and ever will make into profound doubt), or if it's recently been repeatedly blundering into it by sheer stupidity and that's how people noticed the bug at about the same time.

Theorem provers aren't magic, and have bugs just like all other programs. They are tools to help us double-check our reasoning. When you skip the reasoning and ask an AI to "prove" something for you that's over your head, you're entering an adversarial pact with the monkey-pawed Devil of Customer Satisfaction.

 

Great example of why using placeholders like xxx@deleteduser.com is a bad idea.

view more: next ›